A method for identifying coal mine water inrush source based on PCA-ELM model was proposed to improve the identification accuracy of mine water inrush source. The basic control variables were obtained through dimension reduction. Principal component analysis (PCA) was applied to the training set, validation set and test set of the ELM model, and the data samples were used to verify the model. The results show that by simulating the classification process within 10 seconds by the PCA-ELM model,the accuracy of PCA-ELM simulation exercise reaches 100%. The accuracy of the BP neural network simulation only shows 83%, which is far lower than the accuracy of the PCA-ELM model.